Bike helmets: what the safety ratings actually test

Every helmet sold in Europe carries a CE mark and a reference to EN 1078. Nearly every rider takes that to mean the helmet has been tested against the kind of crash they are worried about. It has not. It has passed one narrow, specific, entirely reasonable test, and that test does not include the impact mechanism most strongly associated with the injuries people fear most.

This is not a scandal and it is not an argument against helmets. The standards do the thing they were designed to do, and they do it consistently. It is just that what they measure is much narrower than the label implies, and knowing the gap changes how you choose one.

Here is what the tests actually involve, what falls outside them, which independent ratings fill the gap, and the one variable that matters more than any of it.

What a certification test actually is

Strip away the paperwork and both the European and American standards describe the same laboratory procedure. A headform wearing the helmet is dropped in a guided free fall onto a fixed steel anvil. An accelerometer inside the headform records the deceleration. If the peak value stays under a threshold, on every anvil, in every conditioning state, the helmet passes.

The drop is vertical. The anvil is rigid. The impact is radial — the force arrives along a line through the centre of the headform. Nothing rotates, nothing scrapes, nothing arrives at an angle. One channel of data, one number, pass or fail.

Two consequences follow immediately. The first is that a helmet that passes is not thereby a good helmet; it is a helmet above a floor. The second is that the standards give you no way to tell two passing helmets apart. Both statements are uncontroversial among the people who run these tests.

The impact the standards do not reproduce

Riders very rarely fall straight down onto their heads. They fall forwards and sideways while moving, and the head arrives at a road surface at an angle, with a substantial component of velocity parallel to the ground. That tangential component applies a torque to the head, and the head rotates.

Rotational acceleration of the brain is the loading mode most closely associated with diffuse injury, because it shears tissue across a large volume rather than concentrating force at one point. Laboratory work comparing bicycle helmets under oblique impacts shows that the angled case behaves quite differently from the radial one, and that construction differences which are invisible in a drop test change the rotational loading transmitted to the headform.

When researchers compared helmet performance in tests representing real-world crash scenarios against the scenarios written into the standards, the two did not rank helmets the same way. That is the practical meaning of the gap: a helmet optimised to pass a radial drop test is not necessarily the helmet that performs best in the impact you are likely to have.

The ratings that do test it

The best-documented attempt to close that gap is the STAR protocol developed at Virginia Tech. Rather than a pass or fail, it produces a weighted score.

The method, published in the biomechanics literature, works roughly like this. Impacts are delivered at several head locations and at more than one velocity, onto an angled anvil so that each impact has a tangential component. Both linear and rotational acceleration are recorded. Each test condition is weighted by how frequently that location and severity occur in real cycling crash data, and the weighted injury risks are summed into a single number. Lower is better, which catches people out, and the number is then translated into a star rating.

Two honest caveats. First, STAR is a laboratory ranking. It uses real-world exposure data to decide what to test, but it is not an injury outcome study, and no one has followed riders in five-star helmets to see whether they are hurt less often. Second, a rating system inevitably shapes design towards itself. That is a good thing when the system measures something worth measuring, and it is still worth remembering.

The useful finding buried in this literature is about price. An analysis of the protective performance of inexpensive wholesale helmets found no straightforward relationship between what a helmet costs and how it performs. A cheap helmet that fits your head properly is not a compromise.

Rotational management systems, honestly

Several manufacturers now fit a layer, slip-plane or elastomeric structure intended to let the shell move slightly relative to the head during an oblique impact, reducing the rotational acceleration transmitted.

The independent laboratory evidence is reasonably encouraging and rather more varied than the marketing suggests. Comparative testing of helmets with dedicated rotation-damping systems found reductions in rotational acceleration in oblique impacts, but with substantial differences between designs — the presence of a system is not by itself a guarantee of performance, and some helmets without one tested well.

What does not exist, as far as published research goes, is epidemiological evidence isolating these systems in real crashes. Nobody has run the cohort study. So the honest position is: a laboratory reduction in rotational acceleration is a good reason to prefer a helmet, and it is not the same as a demonstrated reduction in brain injury. Anyone telling you otherwise is ahead of the evidence.

Fit dominates everything above

All of the preceding is about differences between helmets. Fit is a difference between a helmet that is on your head and one that is approximately near it, and the effect size is larger.

A case-control study of children found that those wearing poorly fitted helmets had around twice the risk of head injury compared with those whose helmets fitted correctly. The helmet models were not the variable. The position on the head was.

Shell shape matters here more than brand. Heads differ in the ratio of length to width, and a helmet that is the wrong shape for your skull cannot be made to fit by tightening the cradle — it will sit high, rock on two contact points, and creep backwards over an hour. Try several shapes rather than several price points. If you ride in the dark, the same fitting session is the moment to think about whether the helmet takes a light mount, which riding at night covers in more detail.

Marks, counterfeits and the five-year question

• Check for a real certification mark inside the shell, not just on the box: a CE mark with EN 1078, or the CPSC statement for a US-market helmet. The mark should be moulded or on a permanent label, with a manufacturer name and a date of manufacture.
• Counterfeits exist and they are convincing from the outside. They copy graphics, not construction. The usual tells are a price far below anything legitimate, a vendor you cannot identify, no permanent internal labelling, and foam that feels denser or lighter than it should. Buy from a seller who would still exist if you needed to complain.
• Replace after any impact that crushed the foam. Expanded polystyrene protects by permanently deforming; once it has, that region has spent its capacity. Damage is often invisible under the shell, so the rule is about the event, not the appearance. Research on EPS in helmets shows that impact performance depends closely on the foam's density and its local variation, which is exactly what a crush destroys.
• The five-year replacement interval is manufacturer guidance, not an evidence-based number. There is very little published work on how bicycle helmet EPS ages in normal use, and no study I am aware of establishing a safe service life. Sweat, UV and heat are all plausible degraders, and a helmet left on a car parcel shelf all summer has had a worse life than one kept indoors. Treat five years as a reasonable prompt to look closely at the shell bond, the straps and the cradle, rather than as a deadline with data behind it.

What the outcome evidence says

Separately from all the laboratory work, there is a large epidemiological literature on whether helmet use is associated with fewer head injuries among cyclists who crash. A Cochrane review and a later systematic review and meta-analysis point the same way: substantial reductions in head injury, with the largest effects for the most serious injuries — on the order of roughly half for head injury overall and roughly two thirds for serious head injury, in pooled estimates.

These are observational studies, and observational studies of protective equipment carry known confounding. People who choose to wear helmets differ from people who do not in riding environment, speed, risk tolerance and much else, and the reviews say so themselves. The pooled effect is consistent enough across designs and decades to take seriously; it is not a randomised trial and nobody should present it as one. Whether helmets should be compulsory is a separate argument, about population-level cycling rates and road design, and it is out of scope here.

What this post is about is narrower and more useful: given that you are going to wear one, what should you pick. The answer is a helmet that fits your head shape properly, sits level, is fastened so your scalp moves with it, and — where you have the choice between two that fit — one that has performed well in a protocol which measures rotation. The certification mark tells you it cleared a floor. It does not tell you anything else, and it was never meant to.

The other thing worth remembering is that the crash you are most likely to have is one you could have avoided. Braking technique on a fast descent, position in a bunch and awareness of wheel overlap prevent more head injuries than any liner does — descending faster and safer and group riding skills are both, in a roundabout way, helmet articles. So is riding in the rain, because wet paint and manhole covers put more cyclists on their heads than anything else on a normal week.